Prior-information based super-resolution optical metrology of 2D nanoscale objects

Abstract

Previous work has shown that optical metrology of one-dimensional objects, such as slit width, can achieve improved accuracy by using prior information from similar objects to train the metrology estimator. Here, we demonstrate single-shot optical metrology of nanoscale elliptical particles by analysing their diffraction patterns to retrieve length, width and in-plane orientation using a neural-network estimator trained on prior information from nano-ellipses with varied dimensions and orientations. Fisher-information flow analysis was used to optimise the physical parameters of the metrology apparatus and maximise measurement accuracy. Using a 633 nm laser, we measure the dimensions of elliptical particles with accuracy down to $λ$/128, corresponding to 4.9 nm, and recover their orientation with 5$°$ accuracy. Our results demonstrate the practicality of optical, deep-super-resolution, single-shot, multiparameter measurements of two-dimensional subwavelength objects, with potential relevance to microbiology and nanotechnology applications.